Google has struck a deal to provide the Pentagon with its AI models for classified use, the New York Times reported on Tuesday.
The agreement came despite internal company protests and ongoing disputes over military applications of the technology. The Pentagon has been rapidly securing access to advanced AI through similar deals with major developers since last year.
The deal reportedly expands a $200 million contract signed in 2025, allowing Google’s Gemini AI to operate on classified networks. The tech-giant confirmed the agreement but provided no further details.
“We are proud to be part of a broad consortium… providing AI services in support of national security,” the company’s spokeswoman Jenn Crider said, adding that the company remains committed to ensuring AI is not used for domestic mass surveillance or autonomous weapons without human oversight.
The Information, which first reported the deal, said the Pentagon can now use Google’s AI for “any lawful governmental purpose” – terms that reportedly mirror agreements with other AI developers. US officials say the new agreement is meant to give the military more flexibility and avoid reliance on a single supplier as it integrates AI into sensitive operations such as mission planning and weapons targeting. However, the deal reportedly allows the Pentagon to request changes to Google’s safety filters while making clear the company cannot veto “lawful government” decisions, a provision that has raised concern among critics.
The agreement has triggered backlash inside Google. In an open letter this week, more than 600 employees, including senior staff, urged CEO Sundar Pichai to halt military use of the company’s AI, warning it could be used in “inhumane or extremely harmful ways.” They called for a moratorium on such contracts, greater transparency, and an ethics board to oversee military partnerships.
The deal follows a fallout between the Pentagon and AI startup Anthropic, which earlier this year refused to loosen safeguards on surveillance and autonomous weapons. The Defense Department labeled it a “supply-chain risk,” effectively sidelining it from future contracts. Anthropic has since challenged the decision in court.
Other firms have taken a different approach. OpenAI recently struck its own deal to deploy models on classified Pentagon networks, while Musk’s xAI has integrated its systems into the military’s internal AI platform, GenAI.mil, used by millions of personnel.
The Pentagon’s AI push has sparked protests from tech workers and activists, who warn the technology could be used in military settings without proper accountability. Demonstrators last month gathered outside OpenAI offices with slogans such as “No AI surveillance state” and “Orwell warned us.”
The Pentagon insists it has no plans to use AI for mass domestic surveillance or fully autonomous weapons but continues to push for broad access to advanced systems for “any lawful use.”